Summary of Work Done on Typex Traffic by OKH/In 7/VI, 1939-1941

Declassified this week by the National Archives, this Target Intelligence Committee report details attempts by the German Army High Command to break British TypeX high grade cyphers in the first years of the Second World War.

TypeX was a British cypher system broadly based on the commercial Enigma machine. It featured a total of five drums, two of which were static when set, and three using Enigma stepping. A large number of different drum inserts were produced which could also be reversed when fitted. The reflector was also rewireable. Letters could be shifted to provide upper and lower case, although this feature could have proved to be a potential weakness in the encypherment. Operating procedures were refined, including the introduction of disguised indicators, as the war progressed and lessons learnt from exploitation of procedural errors discovered in enemy signals communication.

Lee Richards
03 March 2018

L.91/6a/3523

TOP SECRET

TICOM/D-83

Attached for your retention is a copy of Ticom/D-83, which summarises the work done by OKH on Typex in the early years of the war as seen from the reports contained in the recently discovered OKH documents.

2. Pendered comments as follows :-

“In reviewing the evidence contained in the file the first point to notice is that OKH were extremely slow off the mark. The initial work by the “Horchleitstelle” was evidently so much time wasted – a simple frequency count of cypher text should have shown them how best to proceed. However, by about May, 1941, OKH were in full possession of all the theory necessary for solving the problem and there can be little doubt that, had they been lucky enough to capture a set of drums at Dunkirk as well as the three machines, the bulk of the Typex traffic up to July 1940 would have been read. All they needed was the drums – and the rest would have followed. Had they succeeded in reading early traffic it seems reasonable to conjecture that they might have maintained continuity beyond the change on 1/7/40 when the “red” drums were introduced. Lack of incentive no doubt accounts for inefficiency in the early stages and by the time the solution of Typex had become a vital problem too much time, which could never be made up, had already been lost. The immunity which Typex enjoyed in the first two years of the war was due partly to the care with which the drums were safeguarded, partly to German inability to grasp the potentialities of the problem, but mostly to sheer good fortune”.

L.91

14th August, 1947.

[Addressed to L with copies to H, H.62, Mr. Alexander, and Dr. Morgan.]

TICOM/D-83

TOP SECRET

SUMMARY OF WORK DONE ON TYPEX TRAFFIC BY OKH/In 7/VI,    1939 - 1941.

The attached is a summary of TICOM document T.2781, consisting of a series of reports by the German Amy Cryptanalytic Agency, OKH/In 7/VI, on their work on British Typex traffic from the beginning of the war until the end of 1941, when the attempt to read Typex was apparently abandoned. The reports are part of the batch of OKH cryptanalytic documents which were uncovered at GLASENBACH in Austria early in 1947 and were recently received from A.S.A, Europe.. This summary comprises all references to work on Typex among these documents. The original German reports, together with a number of tables and diagrams, are available for reference in the TICOM library.

TICOM

8th August, 1947.

No. of Pages:     9

Copy No: 11

Distribution :-

L.S.I.C.

1.           T.

2.           S.

3.           H.

4.           M.

5.           L.

6.           Z.

7.           H.71.

8.           H.62.

9.           H.63.

10-11     L.91

12.         Mr. Alexander.

13.         Dr. Morgan.

14-17.   Ticom Files.

U.S.L.O.              

18.         U.S.L.O.

19-22.   Op-20-2. (via U.S.L.O.)

23-26.   A.S.A. Washington. (via U.S.L.O.)

27.         Chief, A.S.A. Europe. (via U.S.L.O.)

External.

28.         Signals 6, War Office.

29.         Signals 5, Air Ministry.

30.         D.S.D.10., Admiralty.

TOP SECRET

TYPEX REPORTS CONTAINED IN THE FILE OF OKH/In 7/VI.

Reference TICOM T.2781.

1. General Story. The file contains reports on Typex research issued between September 1940 and December 1941. Included also are copies of various documents on Typex which were captured. It is possible from this material to follow fairly closely the progress of the investigations.

2. The first reports on Typex are headed “Chef HNW/Horchleitstelle, Berlin”. The investigations were carried out at the “Horchleitstelle” by a man named BREEDE and apparently lasted over a year. The conclusions reached at the end of them were totally erroneous from start to finish.

3. At the end of September, 1940, the material was handed over (in bits and pieces) to OKH/In 7/IV4. It was soon realised that the job was a major undertaking. Further resources were brought to bear and in January 1941 the work was taken on by In 7/VI. Largely as a result of an elaborate indicator analysis, in May 1941, In 7/VI succeeded in forming an accurate picture of how the machine worked. They had evidently almost completed their research when they found out by chance that OKW/Chi were in possession of three captured machines (without wheels) which had been pinched at least 11 months earlier! A certain amount of progress was made on traffic of March, April and May 1940 but nothing was read. The last report, by In 7/VI, is dated December 1st, 1941, by which time it had been concluded that without a further pinch the machine was insoluble.

4. Details. The first item in the files is a copy of the Typex daily machine setting key for May and June, 1940. It was captured in Belgium and presumed to be the key for the B.E.F. in that country. The second item was probably captured at the same time (May 1940) and is a copy of a letter from C.S.O. G.H.Q. (i.b.iii) to the C.C.O’s G.H.Q., 1 Corps, 2 Corps, 3 Corps, L. of C. It consists largely of an extract from a W.O. letter from J.H. Williams concerning the use of indicators. The W.O. letter expresses concern at the laxness of operators in moving the drums of the machine at the end of one message (or message-part) and the start of the next. In support of this a number of examples are quoted :

“One Message in 10 Sections.

”

It was stated that in these examples the operator had frequently merely moved his drums to the next large letter and sometimes not moved them at all. (It will be recalled that at this time Typex indicators were sent en clair and that only the 14 large letters were permissible).

5. On the basis of these two documents the “Horchleitstelle” reached certain very obvious conclusions which are explained in a report dated 27th September, 1940. These were :-

(a) The machine had 5 wheels.

(b) The last two wheels did not move during encypherment.

Please turn over

2

TOP SECRET

(c) The letters on the wheels were divided into 12 small letters and 14 large letters. [This was already known from an indicator study].

(d) There was a daily machine setting consisting of the wheel-order. [Significance of pin-tyre setting not appreciated].

6. These somewhat unsubtle deductions were correct as far as they went. It is clear from the report, that the 14 large letters were a most baffling feature and tremendous importance was attached to them. The latter part of the report is devoted to an elaborate and fantastic theory concerning the behaviour of the machine. Each wheel was a sort of performing flea which jumped about erratically, but always a multiple of 7, 14 or 21 skips.             This theory was practically proven by the “decryption” of the start of a certain message which read “For information casualty report........” without spaces or punctuation. The general conclusion in this report was that the system was a machine but by no means insoluble.

7. After the September report the Horchleitstelle continued to work on their remarkable theory of the machine. A long report dated Berlin, 26th November, 1940, covers much the same ground as their previous effort. Some of the more romantic passages are underlined in blue. One such remark reads :

“It can be considered as established that the number 14 is the clue to the machine and is fundamental to the system”.

At this point it is hardly surprising that the Horchleitstelle got stuck and the work was handed over to OKH/In 7/IV.

8. It seems probable that In 7/IV could not spare much time for Typex research - at all events the file contains no reference to their efforts. The next item is, in fact, captured R.A.F. “Confidential Communications Orders, 106-117” dated 1st September, 1940. These contain amongst other things general rules for encyphering figures, and instructions for numbering the “Black” and “Red” drums - nothing of any real value to a cryptanalyst.

9. It was some time in January 1941, that Typex research was tackled by the (newly-formed) In 7/VI with the assistance of a Hollerith Section which had been built up “in the meantime”. This time the problem was tackled absolutely from first principles in a thoroughly businesslike way. The results of their efforts are contained in three reports dated 14th May, 1st July and 5th July, 1941. The substance of the first of these reports is all contained in the third and the progress of the investigations can easily be followed from it.

10. Report 1 of 5th July, 1941. The traffic worked on was in the period February-June, 1940 - i.e. it was already about 12 months old. [In July 1940 the distinction between “large” and “small” letters was dropped]. In March, April and May the volume of traffic averaged rather more than 100 per day and it was these three months which were used principally in the investigation. Traffic for February and June was considerably lighter.

11. The importance of dealing with “homogeneous” material was appreciated and, with this object in view, frequencies 6750 kc/s and 13500 kc/s were studied. Call- signs included were (according to the report) :-

During the period under consideration both call-signs and frequencies were fixed. There was a certain amount of additional traffic (e.g. from ZUD Pretoria) which was considered dubious and not included.

12. Most of the messages were single-part and from 50-70 groups long. There were also a number of multi-part messages - notably one of 26th February, 1940, from GFJ2 to ZUD which had 16 parts. This was, in fact, the same message the

Please turn over

3

TOP SECRET

beginning of which had been so successfully “decrypted” by the Horchleitstelle (see para 6).

13. The captured material which was in the hands of In 7/VI at the time has already been described. The most important items were the Typex keys for May and June 1940 (presumably Army keys) and the extract from the W.O. letter (para 4). From this information it was deduced that the machine consisted of five wheels; that wheels IV and V did not move during encypherment; that the “large” letters were arranged alphabetically round the tyres; and that operators were careless about moving the wheels between the end of one message and the start of the next.

14. Starting from the above hypotheses a careful indicator analysis was made. Indicators were first sorted by Hollerith equipment as far as possible into chronological order. Successive indicators were then studied, particularly indicators of multi-part messages. It was known from the captured documents that cypher operators were slack in moving their drums between message parts and that the starting position (as given in the clear indicator) for, say, part 2, would not differ greatly from the finishing position for part 1. For part 1, therefore, the starting position was known (from the part 1 indicator) and the approximate finishing position was also known (from the part 2 indicator). Since the length of the part was also known, it was possible to guess fairly accurately the amount which each wheel had progressed during the encypherment of a given number of letters. From this it was established that the large letters on the tyres were alphabetically arranged; that wheels IV and V did not move during encypherment; and that most probably wheel III moved regularly one step for each encypherment.

15. As regards wheel I, the indicator analysis revealed that it moved, on average, about 4 to 6 steps for 300 to 400 encypherments, but that it apparently moved either forwards or backwards. The results for wheel II at this stage were indeterminate. Further analysis of wheel I, however, showed that on certain days it appeared to move definitely forwards and on other days backwards. In fact, it was established that the forwards and backwards motion of this wheel was dependent solely on the daily key and that 3/5 of the days had a forwards motion while 2/5 had a reverse motion. From this it was argued that, of the five available drums, three were “forwards”‘ drums and two were “backwards” drums and that the forwards and backwards motion of wheel I was determined on any day by which of the five drums happened to occupy the first (left-hand) position in the machine. The simplest explanation of the forwards and backwards motion was that the phenomenon was apparent rather than real and that it was due to the fact that on two of the drums the alphabet was written in the reverse order.

16. If this hypothesis were correct, then the same phenomenon should be observable on the other four wheels. A further, more detailed, analysis for wheels III, IV and V showed that this was, in fact, true. The distribution curves which showed the amounts that each wheel had been moved between the end of one message-part and the start of the next produced very much sharper peaks; the curves also lost their symmetrical shape and indicated that operators preferred to move their wheels forward (i.e. in the direction of the motion) rather than backward. This tendency had already been observed in connection with the Enigma and confirmed that the results as far as they went were correct.

17. An attempt was now made to reconstruct the hatted alphabet of large and small letters round the tyres. In the first place, it was assumed from the captured W.O. letter that the large letters were arranged alphabetically and this assumption was confirmed by the consistent results of the indicator analysis. Secondly, it was assumed that the alphabet would be the same for all wheels. Presumably the small letters (which were not permissible for indicators) filled in the gaps between the large letters. A statistical analysis of message parts with the same starting letter for wheel III and the same length (mod. 26) was

Please turn over

4.

TOP SECRET

carried out and the following arrangement of large letters was deduced :-

A.FJ..KL..N..O.RS.T.VW.XZ.

The dots represented unknown small letters, and it was believed that this alphabet did not differ very much from the real alphabet on the tyres.

18. A further study of the W.O. letter with its examples confirmed the above alphabet and enabled some of the small letters to be placed in the holes :-

A.FJ..KL..N..OpRSqT.VWyXZb

It was now noticed that such small letters as had been placed were also in alphabetical order and from this it was deduced that all the small letters were in their natural order. This gave the alphabet :-

AcFJdeKLghNimOpRSqTuVWyXZb

which was believed to be “highly probable”.

19.  In actual fact one of the hypotheses was wrong, although the Germans never discovered it. The alphabets on the tyres of the original “Black” drums were not all identical but differed from drum to drum. This fact, which might have been of great value to a cryptanalyst, ought to have been discovered by In 7/VI because the examples quoted in the W.O. letter give very strong evidence in favour of it. As events turned out, however, it was not of much consequence, but it is worth mentioning because it is the only mistake in what was otherwise a sound piece of investigation. The alphabet in para 18 which was postulated for all the wheels in fact approximates very closely to the alphabet on wheel II of the Black drums.

20. Using the alphabet they had recovered In 7/VI next proceeded to a more elaborate indicator analysis. They proved beyond doubt that wheel III moved regularly one step for each encypherment. Proceeding now to an examination of the approximate rate of progression of wheel I, it was shown that this wheel on some days progressed “slowly” on some days “fast” and on other days at an intermediate pace. The same was apparently true also of wheel II. Following up this lead it became clear that this was due to the differing number of turnover notches on each drum. Some drums had three notches and some had four. Finally, it was conclusively proved that of the three drums with “forwards” motion one had three notches and two had four, and that the two drums with “backwards” motion both had three notches.

21. The conclusions were fairly obvious - namely, that Typex worked on the same principle as the Enigma with an identical turnover mechanism. The only difference was that the Enigma plugboard was replaced by wheels IV and V in Typex. The daily key consisted merely of wheel-order and pin-tyre settings and the message-setting was simply given as an indicator en clair which, determined the window readings for the initial alignment of the drums. The fact that the turnover notches were in fact cut in the body of the drum and not in the rotatable alphabet tyre was not discovered (and could scarcely have been discovered from the evidence available).

22. One of the results of the indicator analysis was, of course, the partial reconstruction of the drum-order on any particular day. The approximate rate of progression of wheels I and II indicated the number of notches on wheels II and III. Moreover, the apparent direction of motion of all the wheels gave an excellent clue as to which positions in the machine were occupied by “forwards” drums and which by drums with “backwards” motion. In this way the drum-order for most of the days in March, April and May, 1940, was reduced from 120 possibilities to either 4, 8 or 12. There can be little doubt that these results must have been substantially correct.

Please turn over

5.

TOP SECRET

23. Having spent considerable time and energy on attacking the problem from first principles and having reached conclusions whose correctness could not be doubted, In 7/VI were shocked to discover “purely by chance” that three machines without drums had already been in German hands for almost a year. The only information which these machines provided which had not already been discovered was about the nature of the plain language - namely that the machine had an Upper and lower case with case change keys V and Z which were encyphered.

24. The report ends with an appreciation of the possible methods of solution. The only method available if the drum wirings were unknown was believed to be on a depth of 10-20 messages. The likelihood of this occurring was considered to be very small. Accordingly the conclusion was reached that it was imperative that some special effort should be made to pinch a set of drums. It was believed that with the drum-wirings known something might be done, but in this report no methods of solution are indicated.

25. The Crypto Report of 1st July, 1941, gives detailed methods for solving the wheel wirings under certain conditions :-

(a) The wiring of all but one of the wheels is known and

(b) There is a crib (of 100 letters).

The treatment of the problem is purely mathematical and of no particular interest.

26. Crypto Report of 9th July, 1941. This report by In 7/VI gives details of a method of solution given

(a) The wheel-wirings and notch positions and

(b) A crib (of 20 letters or so).

It is, in fact, the well known method of rodding on wheels III, IV and V. It was known to be valid up to the end of June, 1940, (when the “Red” drums were introduced) and it was thought that it would probably work after this date also. The labour required can be briefly described. The maximum number of arrangements of wheels III, IV and V on any day was 5 x 4 x 3 = 60. However, owing to the wheel-order reduction yielded by the indicator analysis the number of wheel-orders would be very much less than 120 - probably not more than about 6 (see para 22). Up to June 30th, 1940, the pin-tyre setting was restricted to the large letters only (this was known from the captured Army keys). The number of possible pin-tyre settings which had to be tried was then 143. The approximate number of tests required would be about 6 x 143 = 16,464. This was not by any means a large number and could certainly be tackled by hand. No fully mechanised method was suggested, but a semi-mechanised scheme using a converted Enigma and a lampboard was suggested. There can be no doubt that it would have worked if the conditions (a) and (b) had ever been fulfilled. Moreover, the step from a semi-mechanised approach to a fully automatic method would not have been a difficult one.

27. Crypto Report of 16th July, 1941. This report deals solely with the stepping progression of the machine and is of little interest. The number and length of the cycles of the machine are calculated under various different conditions.

28. Report 2 of 29th November, 1941. This report gives the results of a Hollerith analysis of cypher text frequencies. The material worked on was again that of March, April and May, 1940, (R.A.F. traffic). The basis of the analysis was the fact that a letter cannot encypher as itself. The commonest P/L letters will thus be the rarest cypher text letters and vice versa. The results of this analysis showed that the two-case feature of the machine was utilised to the full - in particular the space-bar character represented by X.

Please turn over

6.

29. In addition to this the beginnings and ends of all messages were “boiled”. This is a recognised technique for use with a “non crashing” machine and is intended to show up any stereotyped phraseology at the beginnings or ends. It was deduced from the boil that 30-40% of messages began AIRX.... and 20-30% began with HQX... It was also discovered that at the end of messages the space-bar (X) was encyphered to make up the length to an exact multiple of 5. (See Appendix A).

30. Report 3 of 1st December, 1941. This is the last report in the file. It concerns investigations carried out on traffic of September 1940 to see what changes, if any, had been introduced at the same time as the “Red” drum on 1st July, 1940. The volume of traffic was low and results were unconvincing. It was established from the indicator analysis that the distinction between large and small letters had been dropped and that both were now permissible as indicators. The percentage of large and small letters which were used in the indicators showed rather sharp differences between different days. From this it was conjectured that possibly the alphabet arrangement on the new drums was different from the arrangement on the old ones - including also the division into large and small letters, if it still existed. However, shortage of material prevented this hypothesis from being confirmed.

31. At this point the investigations appear to have been concluded. The final remark is :-

“One can hardly expect to make any further progress without another pinch”.

32. The final item in the file consists of correspondence between OKH and the Army Command in France. An attempt to find Typex drums among material captured in France had evidently failed; and an approach to certain French firms which had previously been doing cryptographic work had also proved fruitless.

APPENDIX A

TABLE OF PERCENTAGE LETTER FREQUENCIES FOR FINAL GROUPS

TOP SECRET

TABLE OF PERCENTAGE LETTER FREQUENCIES FOR FIRST GROUPS.